[paynejd]

test_inverse_mappings_are_unchanged (added in #949) failed intermittently on CI: three of five runs on the same tree on 2026-10-22, including both attempts of master's Build at 0c4bae2, which blocked the deploy. It passed locally every time (alone, in its module, and in the full 2,720-test suite).

Cause. The per-concept mapping querysets in ConceptAbstractSerializer.get_mappings and ConceptVersionDetailSerializer.get_mappings have no ORDER BY, and Mapping has no default ordering. The test makes the same request twice. The first request saves checksums on the mapping rows it serializes, which rewrites them, so on a database with earlier tests' leftovers the second request can return a concept's mappings in a different order. The failing reprs had identical lengths, which fits reordered mappings of the same shape. #949's batched path already orders by id.

Fix. Both get_mappings implementations order a concept's mappings by id, for inverse and direct, collection and non-collection. Which mappings are returned doesn't change. The test now compares plain lists, so a failure names the differing element. (response.data is a ReturnList, so assertEqual never printed a diff before.) It also asserts each concept's mappings are id-ordered.

Verified. Locally, 446 tests pass (core.concepts, core.mappings and the concepts integration module) and pylint rates 10.00. CI ran on this PR. The branch's earlier commits added and then removed a temporary diagnostic, so squash-merge it.

Part of OpenConceptLab/ocl_online#456

🤖 Generated with Claude Code

https://claude.ai/code/session_01TjM9dGow2KV1Gwr2dWQdph

[paynejd]

Codex review, pass 1 (codex-cli 0.163.0, commit 155eaf3)

Scope: net diff origin/master...155eaf30b6067ba552bd0c46805dd381604a7bed (the per-concept mapping ordering fix and the test change). Reasoning effort xhigh, read-only.

Verdict: clean. No findings.

findings.json

{
  "reviewer": "codex-cli 0.163.0",
  "perspective": "defects",
  "mode": "full",
  "commit": "155eaf30",
  "verdict": "clean",
  "prior": [],
  "findings": []
}

[paynejd]

Simplification review (fresh Claude subagent, commit 155eaf3)

Verdict: clean, nothing to cut. The change is four order_by('id') calls and a small test tightening; the id-order assertion pins the fix, so it stays. No commits after Codex pass 1.